Large-eddy Simulation of Realistic Gas Turbine Combustors

1 Introduction
The combustion chambers of gas-turbine based propulsion systems involve complex phenomena
such as atomization of liquid fuel jets, evaporation, collision/coalescence of droplets,
and turbulent mixing of fuel and oxidizer giving rise to spray-flames. Accurate observations
and quantitative measurements of these processes in realistic configurations are difficult and
expensive. Better understanding of these flows for design modifications, improvements, and exploring
fundamental physics demands high-fidelity numerical studies in realistic configurations.
Specifically, good predictive capability for swirling, highly turbulent reacting flows in complex
geometries is necessary.
To date the engineering prediction of such flows in realistic configurations has relied predominantly
on the Reynolds-averaged Navier-Stokes equations (RANS) [2, 3]. In RANS, turbulence
models for the Reynolds stress tensor provide time (or ensemble) -averaged solutions to the
Navier-Stokes equations. Though computationally efficient, RANS-based models for two-phase
reacting flows do not represent the relevant flow quantities accurately even in simple configurations.
LES and direct numerical simulation (DNS) techniques have been shown to give good
predictions of turbulent flows in simple configurations [4]. Recently, Pierce & Moin [5] have
shown the superiority of LES to RANS in accurately predicting turbulent mixing and combustion
dynamics in a coaxial combustor geometry. Kim & Syed [6] and Mongia [7] provide a
detailed overview on the importance and role of LES in designing advanced gas-turbine combustors.
The flowfield inside the combustor is highly swirling, separated and turbulent with
complex features such as mixing of secondary cooling air with hot combustion products. The
spray flame is stabilized by the recirculation bubble created by swirling flow. Multiple, turbulent
jets in cross flow play an important role in scalar mixing; may influence pollutant formation,
and elimination of any ‘hot-spots’ in the combustor exit. LES is considered very attractive in
predicting these flow features and their sensitivity to design modifications. However, presently
LES has largely been used to investigate flows in simple configurations. Our goal is to extend
the LES methodology to realistic geometries involving complex physics of multiphase, reacting
2